Gross agricultural output: a quantitative, unified perspective, 1500-1850

Texto

(1)An Agrarian History of Portugal, 1000–2000 Economic Development on the European Frontier Edited by. Dulce Freire Pedro Lains. LEIDEN | BOSTON. For use by the Author only | © 2017 Koninklijke Brill NV.

(2) Contents Acknowledgements ix List of Illustrations x List of Contributors xiii Introduction 1 Dulce Freire and Pedro Lains. part 1 State Formation and Malthusian Growth 1 The Reconquista and Its Legacy, 1000–1348 13 António Henriques Introduction 13 The Reconquista 15. Institutional Developments 24. Cities, Markets and Agro-towns 32. Production and Productivity 35 Conclusion 39 2 The Black Death and Recovery, 1348–1500 45 Ana Maria S.A. Rodrigues Introduction 45. Plague, War and Demographic Decline 46. The Reversal of Fortunes 52 Commercialization 56. The Impact of the Early Overseas Expansion 61 Conclusion 62. For use by the Author only | © 2017 Koninklijke Brill NV.

(3) vi. Contents. Part 2 New Frontiers, Crisis and Growth 3 Coping with Europe and the Empire, 1500–1620 71 Susana Münch Miranda Introduction 71. Population Pressure and Urban Expansion 72. Institutional Framework and Cycles of Agrarian Output 77. The Landscape and the Production Structure 85. The Marketplace 90 Conclusion 93 4 Conflict and Decline, 1620–1703 101 Margarida Sobral Neto Introduction 101. Domestic Settlements and Emigration 102. Colonial Trade 105. Subsistence Farming and the Market 108. Output Cycles 119 Conclusion 126 5 Extensive Growth and Market Expansion, 1703–1820 132 José Vicente Serrão Introduction 132. The Rural Landscape 134. The Macroeconomic Context 138. Grain Issues 147. Property Rights 155 Conclusion 160 6 Gross Agricultural Output: A Quantitative, Unified Perspective, 1500–1850 172 Jaime Reis Introduction 172. Method and Data 175. Trend, Cycles and Short Term Fluctuations 184. For use by the Author only | © 2017 Koninklijke Brill NV.

(4) Contents. vii. Are the Results Consistent? 194 Conclusion 200. Statistical Appendix 201. part 3 Growth, Structural Change and Economic Policy 7 Growth, Institutional Change and Innovation, 1820–1930 219 Amélia Branco and Ester Gomes da Silva Introduction 219. Land, Output and Productivity 220. Institutional Change 228 Innovation 233 Conclusion 239 8 Agricultural Policy, Growth and Demise, 1930–2000 245 Luciano Amaral and Dulce Freire Introduction 245. The Growth of Output and Productivity 247. State Protection 252. The Effects of the Common Agricultural Policy 262. State, the Corporatist System and other Organizations 268 Conclusion 272. Part 4 Lessons from the Second Millennium 9 Agriculture and Economic Development on the European Frontier, 1000–2000 279 Pedro Lains Introduction 279. Lessons from the European Frontier 281. Convergence in European Agriculture 293 Conclusion 307. For use by the Author only | © 2017 Koninklijke Brill NV.

(5) viii. Contents. Appendix: Maps and Figures for Climate, Relief, Administrative Divisions, Urbanization, and Infrastructures in Portugal, 1000–2000 315 Index 338. For use by the Author only | © 2017 Koninklijke Brill NV.

(6) chapter 6. Gross Agricultural Output: A Quantitative, Unified Perspective, 1500–1850 Jaime Reis L’agriculture n’a jamais été bien fleurissante au Portugal. balbi 1822: vol. 1, 143. ⸪ Introduction1 In the current state of knowledge, a quantitative approach to the long-run performance of Portuguese agriculture taken as a whole is possible only from the mid-19th century onward. Thanks to recent research mostly based on published sources, we now have at our disposal continuous time series which span the period from 1850 to the present and cover the main dimensions of this sector. Sufficient evidence has also now been assembled to make possible the estimation, at regular intervals, of the efficiency of factors of production at the aggregate, sector and factor levels.2 For earlier times, however, there is an almost complete dearth in this respect, notwithstanding the considerable volume of micro-oriented scholarship on the medieval and early modern periods, often producing copious and valuable statistical results. The present chapter has two aims. One is to put together, for the years from 1500 to 1850, a similar though less ambitious collection of macro time series, while naturally using different sources and methodologies. The second is to employ this information in order to establish a broad quantitative view of how 1 This study has received valuable financial support from the Fundação para a Ciência e Tecnologia (projects PTDC/HAH/70938/2006) and (PTDC/HIS-HIS/123046/2010). It has also benefited greatly from comments and discussion with António Castro Henriques, Conceição Andrade Martins, Leonor Freire Costa, Carlos Faísca, Nuno Palma and the editors of this volume. 2 For the period 1846–1992 see Lains (2003). The picture of the 20th century can be completed from Soares (2005) and Chapters 7 and 8.. © koninklijke brill nv, leiden, ��7 | doi 10.1163/9789004311527_008. For use by the Author only | © 2017 Koninklijke Brill NV.

(7) Gross Agricultural Output. 173. the agrarian economy evolved from the end of the Middle Ages to the onset of a modern agricultural sector in the 19th century. Pursuing these goals has several benefits. Firstly, it enables us to establish, for the first time, a unified long-run perspective of the dynamics of the entire period, thanks to a new set of consistent quantitative indicators and to the inter-temporal comparisons which they allow. Moreover, it does so over an exceptionally long period of three and a half centuries. Secondly, it delivers a macroeconomic framework into which the findings of the many micro studies, which are the foundation of Early Modern agrarian history, can be incorporated. Since this allows a higher degree of interaction between these explorations, it also considerably raises their respective analytical potential. Thirdly, it offers a number of new quantitative instruments which enable us to sharpen traditional and oftentimes unclear images of both the aggregate performance and the role of agriculture in the overall economy. Finally, it makes greater accuracy possible in assessing derived variables such as per capita food consumption, which are at the heart of current long-term macro debates, in particular those concerning the Malthusian regime. Without such a critical link a model of long-run population size and dynamics in the pre-industrial era cannot be properly established.3 Much has had to be left out of the present analysis. In particular, this is not an exhaustive compilation of all the measurable aspects of the agrarian past. It does not consider subjects like land tenure and access to land, labor relations, crop acreage, or taxation and institutions, which are considered extensively throughout the chapters of this book but at present are still hard to organize into a global vision. Inevitably, the focus here is on agriculture as a whole. Some attention is paid to sub-sectors and products, but the burden of our exercise is for the most part the macroeconomic picture. In addition, it suffers from the limitation of not embracing quantitative trends prior to the early 1500s, a consequence of the still fragmentary and deficient quality of the data available for this period. To complete this overview of the Early Modern period, two further aspects should be taken into account. One is to consider what it was that Portuguese farmers produced during those years. The other is an outline of how agriculture may have evolved during those three and a half centuries, given the constraints and the stimuli to which it was subjected. Natural conditions, in particular agrological and climatic constraints, cultural persistence and the structure of the economy determined a pattern of productive specialization in Portugal which was typical of southern Europe at 3 For a discussion of this problematic in the Portuguese context see Palma and Reis (2016).. For use by the Author only | © 2017 Koninklijke Brill NV.

(8) 174 Table 6.1. 1515 1838 1850. Reis Structure by value of agricultural output (percent). Grain. Wine. Oil. Meat. Total. 38 53 45. 15 36 19. 3 17 5. 44 n.a. 31. 100 100 100. Source: 1515, Godinho 1968: Vol. 2, 20; 1838, Franzini 1843; 1850, Reis 2000: 39.. the time. Output consisted mainly of grain – mostly wheat, maize, rye, millet and barley – animal products, wine, and olive oil, and was almost entirely directed towards feeding the population. Exportables and raw materials (hides, linen, wool) for domestic industry also played a role, albeit a lesser one. It is likely, as far as we know, that early in our period of study (see Table 6.1) the livestock subsector predominated, followed by grain, wine and olive oil in diminishing degrees of importance, a distribution which is hardly surprising. The lingering effects of the Reconquista and of the Black Death left Portugal at the start of the 16th century a thinly populated country (13 to 15 inhabitants per square kilometer in 1527–32) with abundant natural resources. Labor was therefore scarce and the standard of living of this frontier economy was relatively high (Medeiros 1993; Henriques 2015). An analogous agrarian regime has been observed for the same period in the rest of the Iberian Peninsula by Álvarez and Prados (2013). They noted, however, that during the 16th century a sustained increase in population brought on a switch to a new agrarian regime characterized by falling land-labor ratios, a retreat of pastoralism, and an expansion of more intensive crop husbandry. Lower real wages and a reduction in per capita food consumption were the consequences and, as a result, from the 1600s to the 1810s Spain moved towards “a lower path” of growth (Álvarez and Prados 2013: 2). Did Portuguese agriculture follow a comparable course? It seems likely that it did. Table 6.1 certainly reveals a long-run shift in productive specialization between 1515 and 1850 which is commensurate with this. In a period when the population was growing significantly in the face of a by now fixed stock of land, as might be expected grain, wine and oil, all of them labor-intensive products, became much more important (56 to 69 percent of total o utput), and left behind land-intensive animal husbandry, which fell from 44 to 31 percent.4 There are thus signs that Portugal may have experienced the transition from a more 4 Between 1500 and 1800 in Portugal the total land-labor ratio fell by a factor of 2.4, while in Spain it fell by a factor of 2.1. See Malanima (2009: 16).. For use by the Author only | © 2017 Koninklijke Brill NV.

(9) Gross Agricultural Output. 175. extensive to a more intensive agrarian regime which occurred at roughly the same time in Spain, with possibly parallel repercussions in the long-run evolution of its primary sector’s output. The present study depends on the guidelines and benchmarks provided by this “Iberian model”, which it uses to analyze some major questions pertaining to Portugal’s Early Modern agriculture. Chief among them is whether a similar long-run regime shift had the same negative consequences as were experienced over the border. The first section following this introduction shows, in terms of method and data, how the country’s gross annual agricultural output at constant prices may be estimated. The second presents the main results and reveals an overall significant increase in the size of this sector. It also introduces a somewhat revised description of the customary main phases of this expansion. The third discusses the robustness of these findings, and the fourth section draws conclusions.. Method and Data. Method There are essentially two ways of quantifying the gross output of a nation’s agriculture during its pre-statistical period. In the case of countries which have a strong background in quantitative history, this can be done directly by gauging the yearly output of its components at current prices and aggregating them. The real value of this estimate is then obtained by deflating this result with the help of an appropriately weighted index of agricultural prices. These procedures are feasible when sufficient reliable evidence exists for the purpose. To date, this treatment has been possible in three economies – Britain, the Netherlands and Sweden.5 The alternative, indirect approach, which we will follow here, has been used in countries like Italy, Spain and Germany, where quantitative history has not yet matured enough to sustain an effort of direct measurement on such a scale (Álvarez and Prados 2007 and 2013; Pfister 2009; Malanima 2011). An indirect indicator is therefore used, namely the total domestic consumption of agricultural products, which is presumed equal to gross agricultural production.. 5 For Sweden, Edvinsson (2011) and Krantz and Schön (2012); for the Netherlands, van Zanden and van Leeuwen (2012); and for Britain, Broadberry et al., (2015). It should be noted, however, that in the last of these cases, despite the abundance of good data and the use of sophisticated estimation techniques, considerable discrepancies exist between the results obtained. See the discussion and references in Kelly and O’Grada (2013).. For use by the Author only | © 2017 Koninklijke Brill NV.

(10) 176. Reis. This outcome is deflated, as above, to obtain values for this variable at constant prices. Since the second approach is well known, it is presented here only briefly. It is inspired by Wrigley (1985) and has since been significantly revised by Allen (2000) and others. It rests on three main assumptions. One is that the balance of exports and imports of agricultural produce is negligible, a condition for accepting the principle of equality between home production and consumption. Another is that food and raw material consumption can be estimated by means of a demand function for total agricultural production (Qa) given by Qa = Ia Pb Mc N (1) in which I is a proxy for real income per capita, P is the real price index for agricultural products, M is the real price index for other consumer goods and N is total population. The coefficients a, b and c are the respective income, own price, and cross elasticities of demand. In accordance with the SlutskySchultz relation, their sum must be zero. All the prices and wages used to calculate these variables are indicated in grams of silver to ensure comparability. The third assumption is that markets for outputs and for factors of production function well enough to avoid serious allocative distortions. The following comments serve to establish how well these three postulates are met in the Portuguese case. Regarding the first one, it should be noted that Early Modern countries usually conducted a certain amount of foreign trade, much of it in agricultural produce, and cannot therefore be termed “closed economies”. Their exports and imports, however, were seldom large relative to gdp. Exceptionally, in Britain and the Netherlands the difference between the consumption and production of agricultural commodities might at times account for as much as one tenth of national output. Elsewhere, this difference has been deemed barely significant (Allen 2000). Was Portugal also an exception or part of the norm? Traditionally, the view has been that the country suffered chronically from food shortages and had to import large quantities of grain in order to meet the population’s nutritional requirements. According to Godinho (1955: 147), this particular deficit was “one of the most important structural features of Portuguese economic history”.6 This view has been recently tested using data for Portuguese foreign trade for. 6 Author’s own translation. For a recent and rare dissenting voice, which concerns the 18th century only, see Serrão (2005: vol. 1, 172) and Serrão, Chapter 5 in this book.. For use by the Author only | © 2017 Koninklijke Brill NV.

(11) Gross Agricultural Output. 177. several benchmarks between 1550 and 1850 (Costa and Reis 2015). Grain imports and wine and olive oil exports, the main items, were considered for this, the conclusion being that the foreign food balance by value during this period was generally negligible and exceeded the 3 percent mark in only one benchmark, in 1850. It thus seems fair to accept that agricultural consumption is a reasonable proxy for this sector’s output.7 Our second supposition is that the aggregate consumption of the fruits of the land was mainly determined by two factors: the number of consumers (i.e. the population) and their purchasing power. The latter was in turn influenced by their real income, the real costs of food and non-food items, and by the respective propensities to consume them reflected in their demand elasticities. Estimating the various dimensions of this function raises several problems. Possibly the most important of these caveats concerns the calculation of nominal income, from which the real income variable is ultimately derived. Ideally, it should encompass the remuneration, throughout each year, of all productive factors – labor, land and capital. Unfortunately, data for them all are not always easy to come by. In the case of the first, time series for daily rates of pay are readily available for many classes of occupation. On the other hand, average days or hours of work per annum, which undoubtedly varied a good deal over time, as well as across employment categories, are not, and this introduces a significant element of uncertainty in the results. When it comes to land, time series exist too, but covering all forms of land access can be problematic. In the case of income from capital, a lesser share of total income (Malanima 2011), evidence at present is practically unobtainable. Therefore, in keeping with the current practice in this literature, it is left out of our calculation. The solution adopted for estimating nominal income is that suggested by Álvarez and Prados (2013: 9), with some modifications. It involves adding yearly total values of annual wage income and total annual land rent income in 1850, at current prices and in grams of silver. At this date, we assume their respective shares in national income were 75 and 25 percent.8 The two series are calculated from this terminal point, used as a base year, all the way back to 1500 and are combined to obtain the yearly current value estimate for total factor income. The common practice in constructing the wage component of this exercise has been to multiply the daily wage rate (either of skilled or unskilled labor) 7 For an estimation for the 18th century see Serrão, Chapter 5 in this book. 8 Malanima (2011: 178) uses similar proportions for Italy between the 16th and the 18th centuries. Clark (2007: 138) proposes for pre-industrial England 80 and 20 percent.. For use by the Author only | © 2017 Koninklijke Brill NV.

(12) 178. Reis. by the working population and by a fixed number of working days in the year and assume as constant the number of hours in a working day.9 Our approach is akin to this, but more realistic in that it takes into account the simultaneous existence of different labor categories, with different rates of remuneration and labor intensities. In practice, this entails aggregating three classes of earnings: of unskilled laborers, earning the corresponding wage and working an assumed 120 days a year; of skilled urban workers, who received skilled wages and were occupied 250 days a year; and of rural non-agricultural labor (proto-industry, services), whom we presume received an intermediate wage and were active 180 days a year.10 The shares of these categories in the overall labor force are drawn mainly from fiscal and demographic data for the different benchmarks and converted by linear interpolation into annual series.11 Though an improvement, this reformulation of the standard method for deriving labor income is still subject to a degree of bias. In presuming that the annual number of days of labor per worker remained fixed for 350 years, it ignores the possible occurrence of an “industrious revolution” such as has been claimed for many parts of the Early Modern European economy (Angeles 2008; Vries 2008). Moreover, in Portugal a long-term rise in the supply of labor per worker does not seem implausible during this period, given the sustained spread of labor-intensive products like maize (Ribeiro 1986) and wine (Martins 1998), and the productive opportunities offered by imperial expansion (Costa et al. 2015). This intuition can be corroborated with some precision. It is done here by comparing the gross silver values of agricultural output estimated from, respectively, the demand and the supply sides at two distant points in time for which data exist, namely 1515 and 1850.12 The result is that to produce the agricultural output actually obtained in 1850 would have meant increasing the average labor effort per worker by about 150 percent relative to 1515. Clearly a fixed labor input-based methodology would 9. 10 11 12. While Allen (2001) uses skilled wages, Pfister (2012) and Álvarez and Prados (2013) employ unskilled wages, and Malanima (2011) a combination of the two. The total year’s work effort may comprise an average of 250 (Allen 2001; Pfister 2012), 168 (Álvarez and Prados 2013) or 165 days (Malanima 2011). The lengths of these three types of working years are taken from Spain’s (1750) Catasto de Ensenada. See Álvarez and Prados (2013: 7). Details on the estimation of these shares and the respective sources are in Section A.1.3 of the online appendix to Costa et al. (2015). The supply side benchmarks are derived from Godinho (1968) and Reis (2000). The demand side ones are estimated using the demand-for-food function described above.. For use by the Author only | © 2017 Koninklijke Brill NV.

(13) Gross Agricultural Output. 179. entail a substantial under-estimate of actual agricultural output in Portugal for later periods and require a sliding adjustment in the work effort variable. In the absence of direct evidence to enable us to distribute this differential over time, we correct the bias by using an indicator of labor intensity which reflects the annual share of maize, the most diffused labor-intensive crop, in total grain production.13 Putting together an indicator of aggregate rental income is not without its problems either. At this time, most of this revenue came from land, little of which was farmed directly by its lords. Possibly a little less than half of this factor of production was under a regime of commercial tenancy (Monteiro 2005). Typically, leases in this case ran for from three to ten years and provide dependable market-based evidence of the value of the services produced by this resource. The remainder was held under long-term or perpetual emphytheutic agreements, whereby the lord received a fixed fee and the tenant enjoyed an assignable right to the exclusive enjoyment of the fruits of the land (Fonseca and Reis 2011; Costa et al. 2016).14 Although the latter leases were more numerous than the former, the corresponding records are far harder to extract from the sources. We have thus assumed that the rent of the first category of contracts provides a reliable guide to the long-run market price for agricultural land as a whole. This finds justification in the fact that tenant turnover in the second, more rigid category of leases might not be infrequent, given the high mortality among tenants and their successors. Since this could lead to changes in fees at every turn, it is reasonable to suppose that emphytheutic charges were thus likely to vary over time and therefore be quite responsive to market forces, and similarly to commercial ones.15 A second difficulty posed by the implementation of the agricultural output demand function has to do with the value of the elasticities which the model requires. In the absence of historical estimates for them, the literature has drawn inspiration from present day cases of third world economies where 13. 14. 15. This proxy is based on data for grain production in the district of Viseu from the 16th to the 19th centuries (Oliveira 1990 and 2002). Comprehensive, quantitative knowledge about other sectors is patchier, and we assume that, during these three and a half centuries, the rise of “industriousness” was of the same order of magnitude as this one. While the first of these arrangements was employed mostly for larger units of production, the latter corresponded to small or minuscule farms. Contractual forms for accessing land are discussed at length in the first five chapters of this book. For an illustration of how this could happen with similar leases in a 16th-century Spanish context see Drelichman and Gonzalez-Agudo (2014). For a similar point about the flexibility of emphyteutic leases in 17th century Portugal see Amorim (1997) and Neto (2012).. For use by the Author only | © 2017 Koninklijke Brill NV.

(14) 180. Reis. such parameters have been directly estimated and which have similar traits to those of Early Modern economies. There is a fairly wide range from which to choose and which go, for own-price elasticity, from −0.4 to −0.7, and, for income elasticity, from 0.3 to 0.6. Although we find the arguments on this matter by Álvarez and Prados (2013: 3) convincing, we have decided to employ those from Allen (2000), so that we can compare our results with his panel of estimates for agricultural output in ten European countries. We have opted consequently for the set the latter proposes, namely a = 0.5, b = −0.6 and c = 0.1. In implementing this demand model, caution is also needed to make sure that the consumer price index (cpi) employed in it is suited to the context under examination. Allen’s (2001 and 2009) “respectability” basket of goods consumed by a hypothetical mid-18th century worker’s household in Strasbourg is widely used by the current literature on Early Modern European price history.16 In order to adapt it to Portuguese conditions, we have altered the original list of items, as well as some of their quantities, whilst bearing in mind that the global nutritional content is not significantly altered. The Portuguese consumer basket thus defined comprises six food items – bread (made of either wheat or maize), meat, wine (instead of beer), olive oil (instead of butter), eggs, and hens (instead of cheese) – and four non-food items – charcoal, linen cloth, soap and lamp oil.17. 16. 17. We have not adopted the recently devised basket for 17th century Madrid, by Andrés and Lanza (2014), because of what seemed remarkably high figures for meat and wine per capita consumption, respectively 42 kilos and 110 liters a year. The quantities in question correspond to a yearly consumption per adult and are the same as in Allen (2001: 421 and 2009: 36), except in the following cases in which we follow his specific alterations for southern Europe. Grain comprises both wheat and maize, in varying proportions according to their respective changing yearly shares in national output. For this sliding scale see footnote 13. Grain is transformed into bread using the formula calculated by Allen (2001: 418–419). Pulses are not included due to a lack of price information. They are replaced by the equivalent amount of grain (52 liters), which provides a similar quantity of calories. Beer (182 liters) is substituted by 68.25 liters of wine, and 5.2 kg of butter by 5.2 liters of olive oil. Five hens take the place of 5.2 kg of cheese, for which Portuguese price evidence is scant. Given the warmer and longer days characteristic of the European south, annual fuel consumption is reduced from 5 to 2 million btus (British Thermal Units) produced from 90 kg per (for the energy equivalent of charcoal see http://www.extension.iastate.edu/agdm/wholefarm/pdf/c6-88.pdf). We use for other non-agricultural goods the same weights as in Allen (2001), and replace lamp oil with an equal amount (5.2 liters) of olive oil, which performed the same function. The price of soap is proxyed by that of olive oil, which was the principal raw material in the manufacture of this item.. For use by the Author only | © 2017 Koninklijke Brill NV.

(15) Gross Agricultural Output. 181. The last of the three assumptions on which our estimate of agricultural product rests has to do with the extent of the rural population’s engagement in markets. As regards the Early Modern Period, the traditional view in European (and Portuguese) historiography has been that peasants were risk-averse and tended to stay away from the imperfect and risky markets of that era. This made them inclined to self-sufficiency and, consequently, impervious, in their production and consumption decisions, to the stimulus of market forces. If this were so, attempting to model their food consumption patterns on the basis of their reactions to market signals, as we are doing here, would be contradictory and would undermine the aims of this endeavor. A growing current in economic history has argued, however, that such a belief misrepresents the situation. Peasants were not as risk-averse as has been claimed and were often able to work their way round hazards, even using markets as part of their strategy for this. One simple reason is that they lacked sufficient land to sustain their families and meet their obligations to pay dues and taxes in cash. They were thus compelled to enter markets in order to sell at least some of their farm produce and even some of their labor.18 Although the traditional view has enjoyed a strong influence in Portugal too (Serrão 1993: 88), the available evidence increasingly suggests an important economic role for rural markets from 1500, if not earlier. Marques, cited by Santos (1998: 149), has claimed that already in the 16th century a majority of the rural population regularly worked to some extent in return for cash payments, and this continued to prevail at the end of the 18th century. Both in the most and in the least densely populated provinces at this time at least 60 percent of families were dependent for their survival to a significant extent on the sale of their members’ labor (Sousa and Alves 1997; Fonseca and Reis 2011). Revealingly, both Oliveira (1982) and Santos (1998) lay the blame for this at the door of the land scarcity which plagued more than three fourths of rural households and made it unlikely that a regime of economic self-sufficiency might have dominated the Portuguese countryside during those centuries.19 The great abundance of price and wage data which research has recently unveiled 18. 19. For a useful discussion of these issues and evidence on the market involvement of the French peasantry before the 19th century see Hoffman (1996: Chapter 3). Morrison and Snyder (2000: 64) claim that in 18th century France servants and the lower ranks of those in agriculture received half of their income in cash, and that further up the social ladder this share would have been much greater. This share, in the 17th century, was 80–95 percent in the province of Entre-Douro-e-Minho where a quarter of the Portuguese population lived. See Maia (1991), Oliveira (2002) and Mota (2006).. For use by the Author only | © 2017 Koninklijke Brill NV.

(16) 182. Reis. validates the notion that, throughout the period, the majority of the population was connected to some extent with rural or urban markets. Data The long-term series employed here to estimate Portuguese gross agricultural production have been gathered under the auspices of the pwr-Portugal project and can be consulted at its website.20 The sources for both wages and prices refer entirely to Lisbon and its rural hinterland, the latter defined as the territory administered by the episcopal see of the capital city. We thus assume the validity of the principle of “national representativeness” of the data pertaining to the country’s major urban centre and surroundings, a procedure generally adopted by the Early Modern wages and prices literature.21 Several reasons justify this choice here. Portugal is a small country (89,000 square kilometers) and Lisbon is centrally located within it. Although land communications in the Early Modern period were poor, the capital had reasonable links by sea, river and land with many of the country’s regions and their markets (Macedo 1951: 105).22 This would permit the circulation of commodities on a not inconsiderable scale. At least by the late 17th and 18th centuries, the integration of markets for food products was hardly insignificant. Undoubtedly, there was a significant transport and transaction cost premium between Lisbon and its more distant markets, but since improvements in transportation during this period were limited, inter-regional price differentials on this account are likely to have remained more or less constant. In the case of labor mobility, formal and practical restrictions were non-existent and qualitative evidence regarding internal labor movements during this period is plentiful, suggesting the existence of wellintegrated markets for this factor (Silbert 1966; Reis 2005). In common with similar projects, our data come chiefly from the accounts of religious foundations, charitable institutions, royal palaces and municipalities. In the case of commodity prices, they refer almost always to market transactions. In the case of wages, they correspond to employment in either agriculture or the building industry and exclude situations in which there were non-monetary complementary remunerations. In order to standardize results,. 20 21 22. This project is entitled pwr-Portugal: Prices, Wages and Rents in Portugal, 1300–1910 and can be visited at http://pwr-portugal.ics.ul.pt/. For a recent discussion of the pitfalls involved in the cases of Britain and Italy see Malanima (2013). See maps A13, A14 and A15 in the Appendix.. For use by the Author only | © 2017 Koninklijke Brill NV.

(17) 183. Gross Agricultural Output Table 6.2 Years in each century covered by data (percent of total). Wheat Maize Meat Eggs Hens Wine Olive Wages Wages bread bread Oil skilled unskilled 16th c. 94 17th c. 92 18th c. 98 1800–1850 100. na 84 94 100. 78 98 99 100. 33 98 97 100. 60 100 100 100. 62 83 100 78. 65 100 100 100. 66 42 79 98. 75 40 65 100. Source: See text.. all monetary values have been converted into grams of silver and prices have been normalized by the metric system. The coverage afforded by our data compilation displays some variance, with the lowest density of observations occurring in the 16th century, especially during its first half.23 Some linear interpolation has therefore been necessary in order to fill empty cells in the time series. In Table 6.2, we display the percentage of years for which there are data in each century, for the most important variables in use here. As might be expected, coverage rises over time and varies considerably across series. Since workers’ daily rates of pay commonly stayed constant for much longer stretches than prices, interpolation in their case is less problematic. Land rents have attracted comparatively little attention in Portugal from historians of this period, and no doubt for this reason attempts at quantifying them are hard to come by. In the present study, our proxy for the value of the services of land is the total rent generated every year by an invariant set of 32 large commercial estates leased out by a charitable institution in the region of Alentejo between 1595 and 1850 (Santos 2003). We retropolate this index back from 1595 to 1565, using an analogous index for a similar set of estates owned by the hospital of Todos os Santos (All Saints) and located north of Lisbon; and, from then until 1500, employing the series for Spain published by Álvarez and Prados (2013).24 23. 24. The problem of inadequate coverage and the solutions adopted for it – inter- and extrapolation – are common in the historical construction of Early Modern price, wage and rent series. The implication is that the results obtained are more suitable for analyzing the long-term picture than the short-term one. See Zanden and Leeuwen (2012). Between 1595 and 1850, the main Portuguese series, in grams of silver, is consistent not only with the Spanish one by Álvarez and Prados (2013), but also with several other. For use by the Author only | © 2017 Koninklijke Brill NV.

(18) 184. Reis. Until recently, data on the year-to-year variation of the total population of Portugal have been unavailable. The only countrywide demographic information consisted of a few household counts at irregular intervals prior to the end of the 18th century. Thereafter, the two proper population censuses were carried out in 1801, 1841 and 1864 (Rodrigues 2008). This has serious negative implications for the present evaluation, since a linear interpolation between population figures based on such far flung benchmarks is bound to smooth excessively and unrealistically the annual movements of Portugal’s aggregate agricultural output per capita. A recent annual series for the country’s population during the years 1500–1850 (Palma and Reis 2016) is now available, however, and enables us to overcome this difficulty.25 It follows the methodology pioneered by Wrigley and Schofield (1989) and combines information on national population stocks, taken from reliable benchmarks, with yearly flow statistics drawn from a sample of regionally representative parish-level demographic studies.. Trend, Cycles and Short Term Fluctuations. For a long time, the tendency of Portuguese historiography has been to regard pessimistically the performance of Early Modern agriculture.26 This consensus has been constructed on an empirical foundation of impressionistic and occasionally quantitative elements, but has rarely aimed at achieving findings that could be expressed numerically or comparatively. Indeed, until the 1970s, attempts at measuring total production accurately were not a part of the discipline’s program.27 The few historians who contemplated this possibility recognized that the lack of adequate sources rendered this goal unattainable. Writing in the 1960s about agrarian Portugal at the demise of the A ncien. shorter Portuguese series from northern parts of the country collected by Salvado (2010). Moreover, it satisfies the condition that to be accurate it should reflect the principal outputs of the sector (Clark 2002). Grain, olives and sheep are represented but wine is missing from the picture. Nevertheless, its price series tracks that of grain quite closely and our proxy can thus be deemed nationally representative from this point of view. 25 For details of this estimation see the on-line appendix to Palma and Reis (2016). 26 See, however, Serrão, Chapter 5 in this book. 27 In 1951, Macedo (1951) made a list of the time series that would be needed as a foundation for “a clear and precise” study of the agrarian history of Portugal in the 18th century: prices, wages, rents, production areas, yields, taxes, etc. Some of these were published a few years later in Godinho (1955), but even so the kind of agrarian history that Macedo had in mind never took off.. For use by the Author only | © 2017 Koninklijke Brill NV.

(19) Gross Agricultural Output. 185. Régime, Silbert warned that “the application of the rigorous present day methods of economic, social and geographic history [to Portugal] seems to us impossible. To change this would require a miracle” (Silbert 1966: vol. 1, 10).28 At about this time, however, an important change occurred. As a result of the introduction of the study of tithes into this field, an exciting new tool of analysis of the agrarian past. Thanks to the influence of the French Annales School, a flow of new research based on this ecclesiastical source spread rapidly. The result was a significant increase in the quantitative component of the study of agricultural history, as well as a drive in new directions of interpretation.29 A more precise breakdown of the ebb and flow of agricultural activity into phases of expansion, contraction and stagnation was one important contribution of this advance. The possibility of measuring the different rates of change involved in this sub-periodization and of assessing the relative importance of different crops in its long- and medium-term development was another. A third was the potential for aggregating production data into a single representative variable, given the fact that, with this method, agricultural output could be linked to a due which was charged not only on all the fruits of the earth without exception, but also at a universal, easy-to-calculate uniform rate of one tenth. Despite all its advantages, not all that was expected of this methodology was attained. The pitfalls proved to be many. The pronounced local character of this type of research made it hard to construct, with tithes, an empirical foundation for the desired national scope of this exercise. One reason is that combining tithe-based micro indicators to reach a higher level of generality, whether regional or national, was hampered by the lack of adequate benchmarks from which to derive the weights for aggregating them. Secondly, even if nominal output values could have been generated from the merger of disparate tithe series, deflating the result always remained a problem because such studies were often not accompanied by the production of broadly based consumer price indices (cpis), the only ones suited to this purpose. Finally, the frequent patchiness of the sources inhibited the construction of reliable, 28. 29. Similar concerns can be found in Oliveira (1982), Magalhães (1993), Serrão (1993) and Amorim (1997). A recent economic history of Portugal by Costa et al. (2016) reveals the persistence of these concerns as evidenced by the caution with which the authors handle in their narrative the macro-economic indicators required for the Early Modern period. The pioneer of this diffusion was Oliveira (1979). His work has been followed, among others, by Campos (1989), Oliveira (1990 and 2002), Maia (1991), Silva (1994), Amorim (1997), Neto (1997) and Mota (2006). For a summary of tithe-based agrarian studies in Europe see Dodds (2007).. For use by the Author only | © 2017 Koninklijke Brill NV.

(20) 186. Reis. long-term, homogeneous time-series such as one needs in order to analyze agrarian history quantitatively over such long and complex stretches of time.30 Our present effort circumvents these problems by means of the new metric as described above. Its advantage lies in being strictly macroeconomic and grounded in a substantially broader database. The result, in the form of an annual estimate of gross agricultural product at constant prices, is displayed in Figure 6.1.31 To assist the reader in making use of it, three main features should be noted. Firstly, it is expressed in abstract numbers, a consequence of the way in which it is calculated. The implication is that, for each year, its value has no concrete significance in itself and can only serve as a relative magnitude, to be compared with that of any other year of the series. For the base year of 1500 we have chosen a value of 100, all other observations being therefore a multiple of this reference point. In the second place, one should note that it is a consistent evaluation. This means that the values presented throughout the period are the result of applying the same methodology systematically to a homogeneous database. All observations are thus strictly comparable with each other, the same being true for the growth rate of any sub-period within the time span considered. Finally, one must remember that we are dealing here with the aggregate output of the sector, not with its per capita value, two very different concepts which have all too often been used interchangeably and confusingly in the literature. The main significance of this time-series lies in its potential for solving quantitatively some of the key issues and dissipating some of the contradictions which have thrived in the analysis of Portugal’s agrarian past. It allows us to deepen and clarify existing debates, but also to assist in raising hitherto unasked questions and probing new avenues of research. The development of appropriate measures of economic efficiency employing these results is an example of a possible area of innovation in the study of Portuguese agricultural history. An examination of Figure 6.1 brings to light its three main temporal dimensions: the long-run performance over the entire Early Modern period; the cycles of growth, recession and stagnation into which the latter can be broken down; and the short but sharp fluctuations which are characteristic of the economies of this era. The first of these concerns the most critical question in this field: how well did the sector perform between the early 1500s and the 30 31. Chapter 4 by Neto illustrates the analytical difficulties associated with the use of tithes when it comes to estimating national agricultural output. The data from which this figure is derived are represented in tabular form in the statistical appendix to this chapter.. For use by the Author only | © 2017 Koninklijke Brill NV.

(21) 187. Gross Agricultural Output 160 140 120 100 80 60 40 20. 1500 1516 1532 1548 1564 1580 1596 1612 1628 1644 1660 1676 1692 1708 1724 1740 1756 1772 1788 1804 1820 1836. 0. gross agricultural output Linear (gross agricultural output). 6 yr. moving average(gross agricultural output). Figure 6.1 Gross agricultural output (1500 = 100) (1500–1850). mid-19th century onset of a modern industrial economy? The answer can now be easily determined. Over the 350 years in question, the country’s agricultural effort measured in real terms roughly doubled, displaying on trend a cumulative annual growth rate of 0.20 percent.32 Although seemingly modest, it is hardly a trivial figure bearing in mind the length of the time span involved. On the other hand, as might be expected it stands in sharp contrast ro the 1.01 percent achieved during the next 100 years (Lains 2003: 254–257), a plausible difference bearing in mind that we are comparing a pre-capitalist with a modern capitalist economy (Cabral 1976). Most important of all, however, is that it settles once and for all one of the major issues in Portugal’s agrarian history, that of the persistent “decadence” or “stagnation” of this sector. Any claim 32. This growth rate is measured between any pair of points representing centred averages on the linear trend line and not between the end points on the curve. If the latter were used instead, this would yield the much smaller rate of 0.09 percent. An alternative, also used in the growth literature, is to measure this rate between two sufficiently distant peaks or troughs of the curve. Given the difficulty in determining the exact turning points of our curve, we do not use it here.. For use by the Author only | © 2017 Koninklijke Brill NV.

(22) 188. Reis. which depicts the country’s agriculture as structurally incapable of sustained long-term expansion is thus finally dispelled. It is evident from Figure 6.1 that the profile of Portuguese gross agricultural output is far from a straight line. In fact, it can easily be divided into a set of coherent episodes of expansion, contraction or immobility, lasting roughly from a half to a whole century at a time. In the more structured historical narratives currently available, these phases are the backbone of the analytical effort whereby the shifting fortunes of primary production are understood as part of a process. It is therefore important to identify and re-appraise them with care. To do so, we have superimposed on the original time-series a seven-year moving average, which suggests, through a myriad of ups and downs of the output curve, the turning points at which these sub-periods may be held to have started and ended. Obviously, the phases which emerge from this procedure may contain some elements of arbitrariness since their definition depends on the kind of “smoothing” we impose on the data. On the other hand, if other criteria were used, the result would hardly have differed, and thus leaves our conclusions intact.33 With the help of this device, we divide the three and a half centuries laid out before us into six meaningful sub-periods. Although some coincidences can be detected with the current conventional historical wisdom in such matters, some interesting differences come to light. They may imply more than minor historiographical corrections. The 16th century illustrates this well. Generally, it has been perceived as a long and unbroken stretch of moderate growth, described by historians as an era of “agrarian expansion” (Costa et al. 2016: 94), which for some began already in the 1450s (Santos 1998: 154–156), and from which “crises” and setbacks tended to be absent (Gil 1965: 34). On the basis of our new estimate, however, reality appears to have been somewhat different. The long but moderate secular upswing is replaced by three distinct shorter phases, the first of which was a sharp decline of about 40 percent in gross agricultural product occurring between 1500 and the mid-1520s. This was a quite remarkable episode, which has received hardly any attention in the literature.34 The ensuing recovery, which lasted until the late 1560s and returned. 33 34. Tests were carried out with 5-, 10- and 15-year moving averages and at most altered the turning points by two years. One factor which helps account for this downswing is the frequency of bad harvests: one in every decade of this period. The most dramatic of these was the harvest failure of 1521– 22 caused by a severe drought, which afflicted Spain, Portugal and Morocco. Agricultural output fell by a third of the level of the previous decade (Godinho 1963–1965: vol. 2, 278).. For use by the Author only | © 2017 Koninklijke Brill NV.

(23) Gross Agricultural Output. 189. agricultural output to where it had been in 1500, is the second stage. The third carries through from here for the next 100 years as an era of oscillations around a horizontal trend line, which ended in the late 1660s. By this time, agricultural output had still not progressed in any enduring way beyond the level at which our study begins.35 Turning to the 17th century, the standard narrative has long perceived this as a classic age of economic torpor after the good fortune which allegedly stamped the 1500s. As regards agriculture, several recent analyses based on micro-evidence from tithe records have concurred that gross agricultural product in 1700 probably still had not exceeded the level attained in 1600 (Oliveira 1990: 55; Neto 2012: 266 and Costa et al. 2016: 177). Yet, they are also unanimous in asserting that by the last two or three decades of that century the sector’s productive rhythm may have been rising again.36 Divergence exists mainly regarding when protracted stagnation began. Was it in 1600 or 1620?37 The information in Figure 6.1 confirms most of this assessment, but with some adjustment.38 On the one hand, it corroborates two cherished notions: that the “crisis” (as periods without growth have often been designated by Portuguese historians) lasted for approximately a century; and that it came to an end sometime in the 1670s or 1680s. On the other, it shows that this long absence of growth started much earlier – either in the 1570s or the 1580s – some thirty to fifty years sooner than has been thought. In the Early Modern era, the 18th century is undoubtedly the sub-period that has enjoyed the least consensus concerning the performance of agriculture. It. 35 36. 37. 38. For a compilation of “calamities” during the Early Modern period see the specific section in pwr-Portugal entitled “Chronology of calamities”. See Miranda, Chapter 3 in this book. While in an earlier work Serrão (1993) declared it impossible to pronounce on the national agricultural trend of the 17th century, he has recently stated (2005: 47–53) that, at least from the 1670s on, it is clear that output was on the increase. In the case of studies which use similar evidence but have a regional rather than a national focus, patterns diverge somewhat from these national perspectives. In the Entre-Douro-e-Minho province, Capela (1987) has found that local agricultural output was taking off already from 1650. In Beira Litoral, there was a similarly precocious economic recovery, according to Amorim (1997), and the same may have happened in Algarve (Magalhães 2008). The narrative of Chapter 4 in this book, which covers basically the same time span, is somewhat at variance with the present analysis. The principal difference is whether there was a sharp decline in the late 16th and early 17th centuries (our view), and whether 1620– 1670 was a period of stagnation (our view) or one of oscillation around a horizontal trend line.. For use by the Author only | © 2017 Koninklijke Brill NV.

(24) 190. Reis. is also the epoch which has probably been the most thoroughly scrutinized and which has attracted the largest amount of research. Disagreement has arisen mainly over two issues. One of them is whether growth was sustained over the whole century, or was concentrated entirely in the first half of it, with the second witnessing stagnation or even a reversal. The former view has been espoused by Costa et al. (2016: 210) with some caution, by Marques (1973: vol. 1, 519), Serrão (2005: 148–149 and 2009: 48), Oliveira (2007: 256) and Lains (2009: 3, 48). The first has been defended by Justino (1988: 28–30), Pedreira (1994: 386) and Monteiro (1998: 317), mainly on the grounds that the decades after 1750 witnessed a pronounced decline in the grain sub-sector.39 The second problem concerns the net variation in output that may have occurred over the course of the century as a whole. For those who have defended the existence of a continuously rising agricultural cycle, the answer to this, obviously, has to be a secular increase. From the opposing point of view, of those who endorse a downswing after mid-century, a net expansion over these 100 years is not a foregone conclusion. The outcome depends on whether the slow-down in the last fifty years annulled the gains of the preceding halfcentury. Only one author, Pedreira (1994: 389), has so far attempted to quantify this outcome by describing this century as a time of “stagnation”. Our new estimate clarifies these doubts. A glance at Figure 6.1 reveals a very marked contrast between the two halves of this century, with an unmistakable downturn in the early 1750s.40 Moreover, it shows that growth did not start abruptly circa 1700 but constituted, as noted above, a longer trend coming from the 1660s. It accelerated around the turn of the century and ended in the 1750s at a historic peak 50 percent above the figure for 1500. At an annual compound rate of 0.72 percent (measured from 1665 to 1754), this was the longest continuous sweep of agricultural growth in Portugal’s Early Modern period and the best result, in terms of levels, ever achieved prior to 1825.. 39. 40. This point is developed by Lains (2009: 34) on the basis of the real income of fifty families of the high nobility between 1750 and 1820. A similar decline has been detected by Ribeiro (2012) lower down the social scale, for the provincial nobility of the region of Coimbra. Regionally-based studies are present once again in the debate, this time on the side of the retreat of agricultural output after 1750. See Silbert (1966), Magalhães (1988) and Santos (2003). Chapter 5 by Serrão above endorses this outline of the agricultural output performance of the 18th century, but only in terms of the macroeconomic approach we follow here. It claims, on the other hand, that a more broadly based assessment, which also takes into account variables such as crop innovation, adaptability to markets and capacity to feed the population will yield a “better indicator of agricultural progress” (p. 147).. For use by the Author only | © 2017 Koninklijke Brill NV.

(25) Gross Agricultural Output. 191. Figure 6.1 helps illuminate a further aspect of 18th century agricultural history. It shows that the output decline following the early 1750s was significant enough to justify the pessimism of enlightened contemporary opinion, but not sufficient to cancel earlier advances. Although by the next trough, around 1811, gross agricultural product had suffered an overall decline of 40 percent relative to its mid 18th century peak, it was still 20 percent above the previous 17th century low, around 1665. As Serrão (2009 and Chapter 5 above) has suggested, it is as if, during the 1700s, an epochal change had occurred in the nature of Portuguese agriculture, and permanently raised its productive capability to a new plane. The last of the six stages into which Portuguese Early Modern agrarian history can be divided runs from the trough in 1811 until 1850. In contrast with the sub-periods which made up the preceding 300-odd years, it is probably the least known and understood. The chief virtue of our gross agricultural product estimate is that it thus allows us to fill an empty space in the overall trend of agricultural performance between the demise of the Ancien Regime and the middle of the 19th century. For a period generally held to have been economically adverse owing to the regression in manufacturing and trade and to the prevailing climate of social and political turmoil, the achievement of an annual growth rate of 0.7 percent – an overall increase of 30 percent – may come as a surprise. This presents us with the puzzle of what could have caused such a re-awakening of primary production, bringing it back to the historic peak of the 1750s. Possible explanations are: the Liberal implementation of a partial land reform in the 1830s; a hypothetical flight of capital and enterprise from the less promising performance of the trade and industry sectors; and the establishment during the 1810s of a protectionist regime for beleaguered grain producers (Ferreira 1995). The respective impacts of these factors remain, however, to be tested. Over the centuries, rapid and often quite significant oscillations in agricultural production have been a salient feature of pre-industrial economies. These successions of “good” and “bad” harvest years have been a major attraction for agrarian historians and are amply documented in their writings. Portugal, in this respect, is no outlier. Indeed, the record of these fluctuations and of their proximate causes constitutes a substantial part of its Early Modern economic historiography. In these analyses, however, agricultural prices, and not the output variations themselves, have played the leading role in the study of the problematic of short-term output variations.41. 41. For examples see, amongst a great many others, Godinho (1955), Oliveira (2002) and Santos (2003).. For use by the Author only | © 2017 Koninklijke Brill NV.

(26) 192 50 40. Reis 1600. 1700. 1800. 30 20 10 0 –10 –20 –30 –40 1500 1509 1518 1527 1536 1545 1554 1563 1572 1581 1590 1599 1608 1617 1626 1635 1644 1653 1662 1671 1680 1689 1698 1707 1716 1725 1734 1743 1752 1761 1770 1779 1788 1797 1806 1815 1824 1833 1842. –50 Agricultural output, 1500-1850: Deviations from trend. Figure 6.2 Agricultural output, 1500–1850: Deviations from trend (percentage). The third major contribution of the newly estimated index of gross agricultural output is the possibility of deriving a better indicator from it for making cross-temporal comparisons of this sector’s variability. This is done here by taking annual deviations from the long-term trend of gross agricultural product and dividing them by the respective trend values in those years.42 Relative to the conventional price-based approach, the advantage of this is, first of all, that it is based on a direct indicator, rather than a proxy, as has happened until now. The second is that, while our indicator reflects the reaction of the entire sector to shocks, in most studies this is not the case. Usually, only one price or a small handful of disaggregated prices have been employed, and in the latter case these may well follow contradictory paths and be difficult to reconcile. A third is that in price-based evaluations a given year is compared with the preceding one or with the mean of a number of preceding years, a standard of comparison which has quite a different meaning from that provided by a trend. Figure 6.2 displays the deviations of annual agricultural output from trend measured in percentage points.43 It allows us to observe the frequency and 42 43. A Hodrick-Prescott filter for annual data is used for the purpose of calculating these trend values. A similar exercise has been carried out for England before the Industrial Revolution by Kelly and O’Grada (2013). Unfortunately, owing to differences in the methodology used. For use by the Author only | © 2017 Koninklijke Brill NV.

(27) Gross Agricultural Output. 193. intensity of these episodes and brings to light the archetypal unevenness in the performance of this sector. If we assume a drop in output of more than 10 percent as the threshold for “bad” agricultural years, the total of such occurrences for the whole of the Early Modern period comes to a total of twenty-one such events. This suggests that serious food scarcity was taking place on average once every seventeen years, that is, more than once in a generation. It also corroborates the traditional view that this was an epoch in which exposure to hunger and its trail of misfortunes was common. On the other hand, a less obvious fact and one that is hardly remarked upon by the literature is the decline over time of the incidence of bad years, which these data reveal. In the course of the 1500s, fourteen episodes may be classified as “lean”, or worse: 1504, 1513–4, 1521–3, 1532, 1537, 1545, 1552, 1556, 1580, 1589 and 1597. It was also the century in which the only truly catastrophic shock of this kind – a fall in output of more than 30 percent – took place, in 1521 and again 1522. In contrast, during the 17th century bouts of serious hunger were endured only five times – 1603, 1636, 1656, 1662 and 1667 – and in the next century, Portugal was visited only three times by this scourge: in 1710, 1748 and 1783.44 During the first half of the 19th century, there was not a single calamitous agricultural year. One may speculate whether it was the colder and wetter climate of the 1500s which determined this “hunger cycle”, and then the climatic improvement, from about the 1630s, which later had the opposite effect on harvests.45 One may also suppose that it was the secular change in the product mix towards greater diversification which rendered agriculture less susceptible to adverse natural conditions and flattened the peaks and troughs of agriculture’s productive profile.46 Or it could be a combination of both and possibly other factors as well. Whichever it is, the main point to be made here is, once again, that shifting from other tools of analysis to output estimates has two important advantages. It enlarges the information pool at our disposal, thus providing a richer and clearer picture of the past. And it takes us further into our exploration of. to estimate gross agricultural product, their results cannot be compared with those presented here for Portugal. 44 Interestingly, the short-term agricultural crises identified in Chapter 5 of this book match only one of these three episodes of “serious hunger” – the one centred on the year 1710. 45 See figures A1 and A2 in the Appendix. 46 The rising integration of regional and local markets and the gradual reduction of institutional barriers detected by Serrão (2009) for the 17th and 18th centuries may have lessened the oscillations of agricultural prices. We cannot infer from this however that it would have reduced variability in output as well.. For use by the Author only | © 2017 Koninklijke Brill NV.

(28) 194. Reis. the past by raising new questions and problems and creating additional demand for new data and for the methodology to produce them.. Are the Results Consistent?. The importance of the issues raised by the construction, for the first time, of a pre-1850 measure of gross agricultural production is sufficient to justify questions regarding the methodology and sources used for this purpose. Some of them have been pondered already in the methods and data section above, and hopefully dispelled. Others, concerning the historical consistency of the results, have not and need now to be discussed. In this section, we try to verify the credibility of our findings by examining them in the light of the broader picture of European agriculture, as well as by checking their plausibility in terms of relevant domestic indicators. The first step is to establish that our long-term profile of Portuguese agriculture does not imply an unusual growth rate. Allen (2000) provides a convenient standard of comparison in the form of a sample of nine West European countries for which gross agricultural output at constant prices during the period 1500–1800 has been calculated. Table 6.3 displays the percentage increase in their respective agricultural outputs over these three centuries, as well as for Portugal. These data are calculated as the variation between the benchmark values at the extreme points of the distribution, rather than along the trend Table 6.3 Gross agricultural output increase, 1500–1800. Increase (%) Italy Spain France Portugal Austria Germany Belgium Poland Netherlands England. 22 31 33 38 43 53 60 95 119 147. Source: Allen 2000: 17 and text. For use by the Author only | © 2017 Koninklijke Brill NV.

(29) Gross Agricultural Output. 195. line, for reasons of comparability with the standard we have adopted here. The conclusion is that Portugal is hardly out of line with the rest, since it belonged to the majority of nations which managed to achieve only relatively modest secular agricultural growth in this period. This contrasts with England and the Netherlands, which reached the exceptional rates of increase of respectively 95 and 147 percent in the same interval.47 Although we may not have erred as regards the rate of change, a second concern here is whether we may not have over- or undervalued the actual level of output instead. The current dearth of country estimates of agricultural output at current prices for this period makes it difficult to reach comparative conclusions regarding levels, as opposed to rates, of change. An indirect way of approaching this problem is from a Malthusian point of view. In an era such as this one, of scarce food supplies, each national agricultural system had to deliver a minimum per capita nutritional requirement in order to sustain its population. Anything less than this “subsistence” food consumption over a significant stretch of time would have implied an untenable demographic situation, as preventive or positive checks would sooner or later have kicked in. Observed population movement can thus be a way of determining whether our calculation of agricultural output is realistic, not by direct comparison with other national outputs but by analyzing whether historical demographic dynamics corresponded to estimated food availability. Figure 6.3 graphs food supply per capita (drawn from our estimate of agricultural output) in Early Modern Portugal and reveals a significant fall in this variable between 1500 and 1850. A long-term decline was not uncommon in Europe at that time, as Allen (2000: 18) has pointed out: “on most of the continent, [agricultural] output … failed to keep pace with population growth”, particularly during the 16th century.48 Indeed, all nine countries in his sample experienced decreases in food consumption in the range of 10 to 33 percent during these three centuries. In Portugal, however, the drop was considerably greater – 60 percent – yet this happened at a time when population was rising impressively, and by 1800 had reached a level almost three times higher than in 1500. The joint performance of these two variables over such a long time span 47. 48. We are unable here to compare agricultural output levels at different points in time because Allen’s (2000) methodology for estimating this variable differs in important respects from ours. In our case, we assume variability in the annual number of days each worker worked per year, in contrast with his preference for assuming a fixed labor input per capita. This biases upwardly the Portuguese figures for levels. More up-to-date figures for Italy and Spain corroborate this finding. See, respectively, Federico and Malanima (2004) and Álvarez and Prados (2013).. For use by the Author only | © 2017 Koninklijke Brill NV.

(30) 196. Reis 110. 380. 100 330. 90 80. 280. 70 230. 60 50. 180. 40. 130. 30 20 1500 1514 1528 1542 1556 1570 1584 1598 1612 1626 1640 1654 1668 1682 1696 1710 1724 1738 1752 1766 1780 1794 1808 1822 1836 1850. 80 Population (left). Food pc (constant prices) (right). Figure 6.3 Per capita food supply and population in Portugal (1500 = 100) (1500–1850) Source: Palma and Reis 2016; AND text. Notes: FOOD pc = Ratio of gross agricultural product to population; food supply is assumed as equal to agricultural output.. seems like a flagrant contradiction of the workings of the Malthusian model. To solve this puzzle we consider two possibilities. One is that we have underestimated the food supply and therefore the diet consumed per capita, which may have been after all a lot closer to “subsistence”. This would explain the absence of any significant demographic reversion. Alternatively, our calculations might be correct and yet so high initially that even at the lowest subsequent levels of decline it never moved below “subsistence”. We consider two food baskets at current prices and valued in grams of silver, which are accepted as standards of “subsistence” for the Eary Modern period. One is the “barebones” diet proposed by Allen et al. (2012: 873) and inspired by the survival efforts of poorly paid labor during this period.49 The other is Federico and Malanima’s (2004: 445) “minimal subsistence basket for Italy”.50 Both 49. 50. It comprises 165 kilos of maize/wheat, 20 kilos of beans/peas (wheat/maize), 5 kilos of meat and 3 kilos of butter (olive oil) for a year. Items in brackets are the replacements in Portugal for those in the original Northern European list. It includes 50 kilos of wheat, 170 kilos of other cereals (maize), 10 kilos of meat, 3 kilos of olive oil and 80 liters of wine. Firewood is in the original but for the sake of uniformity is ignored here.. For use by the Author only | © 2017 Koninklijke Brill NV.

(31) 197. Gross Agricultural Output 1400 1200. Silver grams. 1000 800 600. Bare Bones Italy PT agric cons pc. 400 200. 1500 1514 1528 1542 1556 1570 1584 1598 1612 1626 1640 1654 1668 1682 1696 1710 1724 1738 1752 1766 1780 1794 1808 1822 1836 1850. 0. Figure 6.4 Subsistence baskets (Italy and “barebones”) and Portuguese per capita food consumption at Portuguese market prices (in silver grams) Source: Italian basket, Federico and Malanima 2004: 445; “barebones” basket, Allen et al., 2012: 873; Portuguese basket, see text.. correspond to the needs of an adult and are converted here using Portuguese prices so as to be comparable with Portuguese per capita food consumption estimates. We cannot use here our earlier quantification of national agricultural consumption (see Figure 6.3) since it is in constant, not nominal silver prices. We obviate this problem by employing a short-cut estimation method proposed by Malanima (2011: 179). This yields an acceptable current price estimate for this variable. It involves multiplying total labor income by a factor of 1.4 to obtain the total income of all production factors.51 Since it is presumed that only a third of the population is economically active (Álvarez and Prados 2007: 326), we multiply this figure by a factor of 0.33 and again by one of 0.6. The latter takes into account Malanima’s additional assumption that food consumption is equal to 60 percent of the value of total income. Figure 6.4 allows us to compare the cost, at Portuguese current prices, of two distinct patterns of “subsistence” food consumption with the value, also at current prices, of the food that, at different points in time, was actually available to the Portuguese population. We conclude that Portugal’s agriculture. 51. The total income from labor in grams of silver is obtained as described above. Malanima (2011: 178) considers that labor income equals 70 percent of total income.. For use by the Author only | © 2017 Koninklijke Brill NV.

(32) 198. Reis. 166. 55. 165.5. 50 45. 165. 40. 164.5. 35. 164. 30 25. 163.5. 20. 163. 15. 162.5. 10. 162 1720 1725 1730 1735 1740 1745 1750 1755 1760 1765 1770 1775 1780 1785 1790 1795 1800 1805 1810 1815 1820 1825 1830 1835 1840 1845 1850. 161.5. 5. heights (left axis). 0. food pc (right axis). Figure 6.5 Average heights and food supply per capita (1720–1850) Source: For average heights, Stolz et al., 2013; for food supply, see Figure 6. Note: Stature measurement taken 20 years after per capita food consumption observed.. consistently provided a level of consumption which was close to accepted subsistence requirements and thus made it possible for the population to withstand Malthusian pressures and keep on expanding. We may thus conclude that our findings are not an underestimate of gross agricultural output. Per capita food intake offers us a second opportunity to corroborate our initial output findings, this time using evidence from the field of anthropometric history.52 A recent study regarding the evolution of the height of Portuguese males between 1720 and 1980 has shown that the standard of living and the level of human capital of this population were, in the long run, the principal determinants of stature (Stolz et al. 2013). The first of these was measured by these authors using the real wage level, but in the present instance we employ average real food consumption instead. This choice is, in fact, an even better solution since we use a variable which has a more direct impact on corporal growth than wages, only part of which are spent on the nutrients which promote physical development. On the other hand, it has the disadvantage for us of covering only the years between 1720 and 1850. Even so, setting these two variables side by side, as in Figure 6.5, provides encouraging support for our estimates of the agricultural trend during the last 52. The link between food consumption and heights has been explored in the same way for Italy by Federico and Malanima (2004).. For use by the Author only | © 2017 Koninklijke Brill NV.